In practice, existing Cat.5e cabling can, on some or all cabling runs, INCIDENTALLY either pass Cat.6 specifications or at least surpass minimal Cat.5e specifications enough to behave within the tolerance margins of 10GBase-T.
The practical limits are dependent on things like when the cable was made - older machinery to make the cable might have been built to only just meet the accuracy needed for Cat.5e, while newer cable might have been produced on the same machinery as the Cat.6a cable and merely labelled differently, either because there was demand for Cat.5e or because cable that marginally fails Cat.6a but meets Cat.5e specs was manufactured due to machine/operator wear/error/maladjustment/contamination...
Also, if a given run works or not can depend on the hardware at both ends - one transceiver design (in a NIC or switch) might be able to just flawlessly work with a marginal cable while another will not.
Small factors like materials aging, someone bending or rolling up the cable, temperature and humidity, someone placing metallic objects too near it ... could be sufficient to turn a working cable run into a non working one suddenly.
Also, the cable could work but create unacceptable EMI problems.
10GBase-T on Cat.5e seems actually common practice - though usually not at maximum length.